Web publication manifest have quite ambiguous author field.

Author term is usually used to mention someone's who've originated the written creative work.
Creator is the wider term. And for case when concept set and sub-set are defined on the same level of bib description its vague which field to use (or fill both records the same).
Lewis Carroll is both: author and creator of "Alice's Adventures in Wonderland" work and only example in WPM body is highlighting that author field should be used. With no clues about criteria of use or explicit difference between 'author' or 'creator'.

The short solution i see in using better definitions of author and creator fields e.g. from LC relator codes vocabulary (text and links below).

The best solution i see is to remove author field. Because even with clear criteria of difference there will be a second level of problem: how to interpret this fields during translation to other standards and forms (which field should be primary if you have display space to show only one and so on).

Below i providing related fragments from neighbour standards and some explanation about the things behind, and why it will be better to remove author term and not to remove creator.

Definitions from W3C Web Publication Manifest

• author - The author of the publication. | One or more Person and/or Organization. | Array of Entities | author (CreativeWork)
• creator - The creator of the publication. | One or more Person and/or Organization. | Array of Entities | creator (CreativeWork)

Definitions from DCMI

• dcterms:creator dc:creator - An entity responsible for making the resource.

DCMI standard don't introduce theAuthor term.

FOAF

dct:creator/dct:agent and maker terms relationship note worth mentioning:

The Dublin Core specification provides term definitions that focus on issues of resource discovery, document description and related concepts useful for cultural heritage and digital library applications. FOAF can be used alongside any variants of Dublin Core, but works most effectively with the most modern Dublin Core terms namespace. Note that here we use the prefix 'dct:' to stand for the DC Terms namespace; however it is not unusual to see 'dc' also used.

• dct:Agent - Dublin Core's notion of Agent is much like FOAF's; Dublin Core says "A resource that acts or has the power to act.", we say "things that do stuff". As nobody has provided a counter-example of something fitting one definition but not the other, we say here that foaf:Agent stands in an 'equivalent class' relationship to dct:Agent (and vice-versa).
• dct:creator - The notion of 'creator' in the latest versions of Dublin Core matches FOAF's notion of 'maker'; based on their definitions, every pair of things that are related by one of those properties are also related by the other. We express this by saying that these properties stand in an 'equivalent property' relationship' to one another.

Definitions from MARC21 relator terms vocabulary and LC LD relator terms vocabulary

• Author aut - A person, family, or organization responsible for creating a work that is primarily textual in content, regardless of media type (e.g., printed text, spoken word, electronic text, tactile text) or genre (e.g., poems, novels, screenplays, blogs). Use also for persons, etc., creating a new work by paraphrasing, rewriting, or adapting works by another creator such that the modification has substantially changed the nature and content of the original or changed the medium of expression
• Creator cre: A person or organization responsible for the intellectual or artistic content of a resource.

By fact, i don't see that aut/cre codes are really common in bib catalogues, usually they omited and only name/auth code defined. And kind of relationship is defined by bibliographers for more specific relation than being originator. So even in the official LC example of MARC21 100 main name entry field you may see no authors or creators being defined explicitly.

I'm pretty sure that its not an occasional detail because bibliographer always holds part of authority institution responsibility and guarantees about single, as complete and correct as possible description being defined in scope of specific functional requirements and description standard. Any case of collision between two actual descriptions usually means that one of the versions should be considered outdated.

This simple authority control rule is main pillar of de-duplication of metadata being possible, and also preventing lot of holywars about which record is right, and question will be which of them is outdated and which should be corrected. The sane answer is the deletion of record with more recent (less background and sync history) control code and maybe its correction toward other record.

All this usually preventing cataloguers from dealing with ambiguous forms of definitions that have a chance to be catalogued differently, even if there are some in the description standard vocabulary.

BibFrame2

During drift of MARC family of standards toward LD its evolved to BibFrame2 (conceptually FOAF-like agent-activity-entity) model.

Bibframe have a complex work concept levels spine with tree levels. And previous version had two levels with different names. Its hard to say that BF2 is clear or stable but its "good enough to use". Due not being mature currently there is a lot of approaches to practical use and entity linking. But the core idea about contributors is something like following::

      (`Work`)        ->         (`Instance`)         ->       (`Item`)
         ^                             ^                           ^
         ^                 (`ProvisioningActivity`)                ^
         ^                             ^                           ^
  ('Contribution`)              ('Contribution`)            ('Contribution`)
      ^     ^                ^    ^          ^    ^             ^      ^
 (`Agent`) (`Role`) | (`Agent`)(`Role`) (`Agent`)(`Role`)  (`Agent`)(`Role`) 
                    |
 WPM author/creator |                      WPM contributor

TheAgent is playing Role (matching relator terms vocabulary) and through this making the Contribution.

Web Publication manifest author/creator fields will have form of Agents making contribution as well as Web Publication manifest contributors

• Contribution - Agent and its Role in relation to the resource. Used with Work, Instance or Item
• Role - Function played or provided by a contributor, e.g., author, illustrator, etc.
• Agent - Entity having a role in a resource, such as a person or organization.
• ProvisionActivity - Information about the agent or place relating to the publication, printing, distribution, issue, release, or production of a resource.

I do not see BF2 as some perfect extension for WPM because WPM seems to be initially flat by design. In opposite BF2 is designed with large temporal dimension and able to describe historical process that came up with object of culture possible to interact directly. BF2 records web is definitely hostile to any table-form representation and actually (e.g. by exclusion of shelf numbers and other storage identifiers space) do not tend to describe final layer of items or digital objects. And even maintainers offical converors from other standards just ignoring information like physical items storage marks or digital publication containers processing details.

In short: BF2 don't designed to provide both isolated and understandable records and merely targeted publishers

AACR/ISFB/FRBR (It's deep domain, so i'm not providing detailed links)

• "author" term is used merely. And usually its the "Work creator"

This standards families are the real ground for all standards mentioned above and DCMI may be considered as the robust and consolidated shortcut down to the bibliographic domain regulation and experience.